U.S. patent number 7,749,171 [Application Number 11/076,789] was granted by the patent office on 2010-07-06 for method for automated detection of a-waves.
This patent grant is currently assigned to NeuroMetrix, Inc.. Invention is credited to Shai N. Gozani, Xuan Kong.
United States Patent |
7,749,171 |
Gozani , et al. |
July 6, 2010 |
Method for automated detection of A-waves
Abstract
In one form of the present invention, there is provided a method
for detecting an A-wave, the method comprising: applying a series
of stimuli to a nerve; recording a series of late responses;
creation of a feature space map from an ensemble of late responses;
identification of clusters within the feature space that represent
A-wave components; consolidation of A-wave components into a
discrete collection of A-waves; removal of false positive A-waves;
and extraction of A-wave characteristics. In another form of the
present invention, there is provided a system for detecting an
A-wave comprising: a stimulation electrode; a stimulation circuit
connected to the stimulation electrode for applying a series of
stimuli to a nerve; a detection electrode; a detection circuit
connected to the detection electrode; and an analyzer connected to
the detection electrode and adapted to detect an A-wave by:
recording a series of late responses detected by the detection
circuit; creation of a feature space map from an ensemble of late
responses; identification of clusters within the feature space map
that represent A-wave components; consolidation of A-wave
components into a discrete collection of A-waves; removal of false
positive A-waves; and extraction of A-wave characteristics.
Inventors: |
Gozani; Shai N. (Brookline,
MA), Kong; Xuan (Acton, MA) |
Assignee: |
NeuroMetrix, Inc. (Waltham,
MA)
|
Family
ID: |
34976244 |
Appl.
No.: |
11/076,789 |
Filed: |
March 9, 2005 |
Prior Publication Data
|
|
|
|
Document
Identifier |
Publication Date |
|
US 20060020222 A1 |
Jan 26, 2006 |
|
Related U.S. Patent Documents
|
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
Issue Date |
|
|
60551556 |
Mar 9, 2004 |
|
|
|
|
Current U.S.
Class: |
600/554;
600/546 |
Current CPC
Class: |
A61B
5/24 (20210101) |
Current International
Class: |
A61B
5/05 (20060101) |
Field of
Search: |
;600/373,378,544-546,554 |
References Cited
[Referenced By]
U.S. Patent Documents
Other References
Bischoff et al., "Significance of A-waves recorded in routine motor
nerve conduction studies," Clin Neurophys 1996; 101:528-533. cited
by examiner.
|
Primary Examiner: Hindenburg; Max
Assistant Examiner: Towa; Rene
Attorney, Agent or Firm: Pandiscio & Pandiscio
Parent Case Text
REFERENCE TO PENDING PRIOR PATENT APPLICATION
This patent application claims benefit of prior U.S. Provisional
Patent Application Ser. No. 60/551,556, filed Mar. 9, 2004 by Shai
Gozani et al. for METHOD FOR AUTOMATED DETECTION OF A-WAVES.
The above-identified patent application is hereby incorporated
herein by reference.
Claims
What is claimed is:
1. A method for detecting an A-wave, the method comprising:
applying a series of stimuli to a nerve; recording a series of
evoked bioelectrical responses; identifying one or more attributes
in each of the responses; utilizing each of the one or more
attributes of each of the responses to create a new data set;
creating a search window that is smaller than the new data set;
searching the new data set for trends by: sequentially applying the
search window across the new data set; and analyzing the search
window by counting the number of data points contained within the
search window at each separate, sequential search window position;
analyzing the new data set by: registering a positive for an A-wave
component if the number of data points contained within the search
window at each separate, sequential search window position exceeds
a predetermined threshold; and registering a negative for an A-wave
component if the number of data points contained within the search
window at each separate, sequential search window position does not
exceed a predetermined threshold; and consolidating the A-wave
components into a single A-wave.
2. A method according to claim 1 wherein the search window has an
area such that the height of the search window is proportional to
the amplitude location within the new data set.
3. A method according to claim 1 wherein the one or more attributes
comprise the amplitude of local maxima and/or local minima.
4. A method according to claim 1 wherein the one or more attributes
comprise the absolute value amplitude.
5. A method according to claim 1 wherein the one or more attributes
comprise the second derivative of amplitude.
Description
BACKGROUND OF THE INVENTION
A nerve conduction study (NCS) is a diagnostic procedure whereby
peripheral nerves are stimulated electrically and then
bioelectrical potentials are recorded from the same nerve at a
second location or from a muscle innervated by the activated nerve.
A nerve conduction study often consists of early and late
potentials. The former reflect direct conduction from the site of
stimulation to the site of recording. Late potentials represent
conduction from the site of stimulation antidromically towards the
spinal cord--reflection along the way or in the spinal cord--and
then conduction back down to the recording site.
The two most common types of late potentials associated with
recording from a muscle innervated by the stimulated nerve are
F-waves and A-waves. F-waves waves are highly variable waveforms
that are caused by motor neuron back-firing and are generally
recorded in all nerve conduction studies--whether pathology exists
or not. A-waves, by contrast, have nearly constant latency and
morphology and are generally not found in the absence of pathology.
Thus, their presence is strongly suggestive of a focal or
generalized neuropathy. The pathological entities most commonly
associated with A-waves are polyneuropathies, particularly
inflammatory neuropathies such as Guillain-Barre syndrome and
lumbosacral nerve root compression.
One class of A-waves, called axon reflexes, are thought to be
generated by collateral sprouting, have a simple morphology and are
usually eliminated by supra-maximal stimulation. A-waves that
persist with supramaximal stimulation, especially in multiple
nerves, are sensitive indicators of electrophysiological
abnormalities.
In the prior art, A-waves are identified using manual inspection of
evoked late responses acquired during a nerve conduction study.
Typically, a clinician views an ensemble of late responses in a
raster format and makes a subjective determination as to whether an
A-wave exists. This approach has several significant
deficiencies:
(i) the subjective A-wave identification process is time-consuming
and may not be performed because of time and resource
limitations;
(ii) subjective A-wave identification does not support
standardization of A-wave characteristics and thus may lead to wide
differences in clinical results; and
(iii) subjective A-wave processing is, realistically, restricted to
identification of the presence or absence of an A-wave--other
A-wave features that may be of diagnostic value are unlikely to be
extracted in reliable manner.
SUMMARY OF THE INVENTION
In response to the deficiencies and limitations of the prior art,
we have developed an automated A-wave detection algorithm which is
advantageous because it:
(i) eliminates the need for the tedious, inefficient and error
prone process of manual A-wave identification;
(ii) ensures consistent A-wave features, thus providing
standardization of A-wave characteristics across electrodiagnostic
studies--including those performed in multiple sites by different
clinicians; and
(iii) supports the automated extraction of a series of A-wave
characteristics that maximize A-wave diagnostic utility.
In one form of the present invention, there is provided a method
for detecting an A-wave, the method comprising:
applying a series of stimuli to a nerve;
recording a series of late responses;
creation of a feature space map from an ensemble of late
responses;
identification of clusters within the feature space that represent
A-wave components;
consolidation of A-wave components into a discrete collection of
A-waves;
removal of false positive A-waves; and
extraction of A-wave characteristics.
In another form of the present invention, there is provided a
method for detecting an A-wave, the method comprising:
applying a series of stimuli to a nerve;
recording a series of evoked bioelectrical responses;
identifying one or more attributes in each of the responses;
utilizing each of the one or more attributes of each of the
responses to create a new data set;
creating a search window that is smaller than a the new data
set;
searching the new data set for trends by: sequentially applying the
search window across the new data set; and analyzing the search
window by counting the number of data points contained within the
search window at each separate, sequential search window
position;
analyzing the new data set by: registering a positive for an A-wave
component if the number of data points contained within the search
window at each separate, sequential search window position exceeds
a predetermined threshold; and registering a negative for an A-wave
component if the number of data points contained within the search
window at each separate, sequential search window position does not
exceed a predetermined threshold; and
consolidating the A-wave components into a single A-wave.
In another form of the present invention, there is provided a
method for detecting an A-wave, the method comprising:
applying a series of stimuli to a nerve;
recording a series of evoked bioelectrical responses;
identifying one or more attributes in each of the responses;
utilizing each of the one or more attributes of each of the
responses to create a new data set;
identifying trends in the new data set; and
analyzing the trends to identify the A-wave.
In another form of the present invention, there is provided a
method for diagnosing a disorder in a patient comprising:
detecting an A-wave in a patient by: applying a series of stimuli
to a nerve; recording a series of late responses; creation of a
feature space map from an ensemble of late responses;
identification of clusters within the feature space map that
represent A-wave components; consolidation of A-wave components
into a discrete collection of A-waves; removal of false positive
A-waves; and extraction of A-wave characteristics; and
comparing the A-wave of the patient with the A-wave of known
disorder.
In another form of the present invention, there is provided a
system for detecting an A-wave comprising:
a stimulation electrode;
a stimulation circuit connected to the stimulation electrode for
applying a series of stimuli to a nerve;
a detection electrode;
a detection circuit connected to the detection electrode; and
an analyzer connected to the detection electrode and adapted to
detect an A-wave by: recording a series of late responses detected
by the detection circuit; creation of a feature space map from an
ensemble of late responses; identification of clusters within the
feature space map that represent A-wave components; consolidation
of A-wave components into a discrete collection of A-waves; removal
of false positive A-waves; and extraction of A-wave
characteristics.
BRIEF DESCRIPTION OF THE DRAWINGS
These and other objects and features of the present invention will
be more fully disclosed or rendered obvious by the following
detailed description of the preferred embodiments of the invention,
which is to be considered together with the accompanying drawings
wherein like numbers refer to like parts and further wherein:
FIG. 1 is schematic drawing showing the translation of an ensemble
of late response into a feature space;
FIG. 2 is schematic drawing showing one preferred translation
function for translating a feature of a late response into a
feature space;
FIG. 3 is schematic drawing showing how the feature space may be
searched for clusters of points that are likely to represent A-wave
components;
FIG. 4 is schematic drawing showing how a positive A-wave component
followed a short time later by a negative A-wave component is
likely to indicate two phases of the same A-wave waveform.
DETAILED DESCRIPTION THE PREFERRED EMBODIMENT
Electrical stimulation of many peripheral nerves (e.g., median,
ulnar, peroneal and tibial) evokes a "late response". This response
is characterized by antidromic (or retrograde) conduction of the
evoked impulse from the point of stimulation to the spinal cord,
reflection of the impulse in a subset of the motor neurons, and
orthodromic conduction of the impulse to a location on that nerve
or to the muscle(s) innervated by the nerve.
The late response may have several components, the two most
commonly associated with muscle recordings are F-waves and A-waves.
The F-wave represents motor neuron backfiring. The A-wave is
generated by pathophysiological processes that are located distal
to the motor neurons. These pathophysiological entities cause
reflection or return of the antidromic impulses prior to their
arrival at the motor neuron.
The present invention comprises a novel system for the automated
detection of A-waves. Among other things, the novel system utilizes
a unique A-wave detection algorithm which will hereinafter be
discussed in detail.
More particularly, the novel A-wave detection algorithm employs a
feature space representation of an ensemble of late responses to
identify A-waves. The preferred embodiment of the algorithm
comprises five steps:
(i) creation of a feature space map from an ensemble of late
responses;
(ii) identification of clusters within the feature space that
represent likely A-wave components;
(iii) consolidation of likely A-wave components into a discrete
collection of likely A-waves;
(iv) removal of false positive A-waves; and
(v) extraction of A-wave characteristics.
In the first step, an ensemble of late responses is mapped into a
feature space. The translation of an ensemble of late responses
into a feature space is shown in FIG. 1. The late response traces
10 (created by a series of electrical stimuli applied to a
patient's peripheral nerve, with the patient responses recorded as
a series of traces by detection electrodes) may consist of F-waves
12 and A-waves 14. A translation function 20 is applied to the late
responses 10, yielding a feature space map 30 consisting of
discrete points 32, each of which represents one feature of one
late response trace. The number of discrete points 32 within the
feature space 30 can be less than, equal to, or greater than the
number of late response traces 10, depending on the number of
features identified per trace.
The feature space map may be created from one or more of the
various features associated with the late response traces. More
particularly, in the preferred embodiment of the present invention,
the feature space map is created using every local maxima and local
minima within each late response trace. Thus, the translation
function 20 of the preferred embodiment is shown schematically in
FIG. 2. The function 20 identifies every local maxima 50 and local
minima 52 within each late response trace 40 (this late response
trace would be one of the ensemble of traces seen in 10). A single
point 46 within the feature space 30 is then created to correspond
to each such maxima 50 or minima 52, whereby the location of the
point 46 is determined by its time of occurrence 42 and its
amplitude 44.
In additional embodiments of the present invention, other late
response trace characteristics and attributes may be used to
construct the feature space. By way of example but not limitation,
such characteristics may include the magnitude of local maxima and
local minima of various linear and non-linear translations of the
response trace 40, including its first derivative, its second
derivative, its absolute value, and its second power. Although, in
the preferred embodiment, the attribute of the maxima or minima
that is mapped into the feature space is its amplitude, other
attributes could be utilized. By way of example but not limitation,
such attributes include the absolute value and second power of the
amplitude. In the embodiment described above, the feature space
consists of two dimensions: time 42 of maxima 50 or minima 52
occurrence; and its corresponding amplitude 44. Higher dimensional
feature spaces have been contemplated and should be viewed as part
of the present invention. As an example, in one such embodiment, a
third dimension representing the magnitude of the second derivative
(i.e., "sharpness") at the time of occurrence of the maxima 50 or
minima 52 is incorporated.
In the second step of the preferred algorithm, the feature space is
searched for clusters of points that are likely to represent A-wave
components. A-waves are defined by nearly constant waveform
morphology and latency. In a preferred embodiment of the algorithm
shown in FIG. 3, a search window 60 of a predetermined temporal
width 62 and amplitude height 64 is sequentially applied across the
entire feature space 30. Any location whereby the search window
overlaps at least a predetermined number of points 70 is registered
as an A-wave component. In a preferred embodiment, the number of
points is defined as a percentage of the number of late response
traces 10. For example, the frequency of A-waves within late
response ensemble 10 can vary from as low as 40% of traces to 100%
of traces, and thus the search window 60 must overlap a number of
points which is 40% of the number of late response traces. In
another embodiment of the present invention, the search window 60
does not have to be a fixed size but can increase in either width
62 (temporal dimension) or height 64 (amplitude dimension) at
different parts of the feature space. For example, as the amplitude
of the features increase, there is a greater variation in the
points. As a result, it is advantageous to make the height 64 of
the search window 60 proportional to the amplitude location within
the feature space 30.
In the third step of the preferred algorithm, the identified A-wave
components 70 are consolidated into A-waves. This is accomplished
by merging A-wave components that are likely to represent different
elements of the same A-wave. For example, as shown in FIG. 4, a
positive A-wave component 80 followed a short time later by a
negative A-wave component 82 is likely to indicate two phases of
the same A-wave waveform 84. In this situation, the components are
consolidated into a single A-wave.
In the fourth step of the preferred algorithm, the specificity of
the ensemble of consolidated A-waves is optimized by applying a set
of heuristic rules. The purpose of this step is to identify and
remove "false positive" A-waves. These are segments of the late
response trace that were identified as A-waves by steps 1-3 of the
preferred algorithm but do not actually represent physiologically
realistic A-waves. In the preferred embodiment, the rules are
predetermined and include, by way of example but not limitation,
minimum amplitude, minimum time of occurrence and minimum
"sharpness". These rules can also be combined. For example, the
minimum amplitude of an A-wave may be defined as a function of its
time of occurrence, whereby A-waves that occur early in the late
response must have a larger amplitude than those that occur
later.
In the fifth step of the preferred algorithm, the final reduced set
of A-waves is analyzed and each A-wave is characterized by a set of
features. One standard feature is the amplitude of the A-wave. By
way of example but not limitation, other features include the
complexity of the A-wave--which may be estimated by the number of
phases in the A-wave, the time of occurrence of the A-wave, the
temporal dispersion of the A-wave, and the persistence of the
A-wave--defined as the percentage of late response traces in which
the A-wave occurs.
The presence or absence of A-waves in a late response ensemble, as
well as the characteristics of these A-waves, can be used as is
known in the art to diagnose neuropathic conditions. For example,
the presence of an A-wave in the peroneal nerve is suggestive of a
chronic lesion of the L5 nerve root, otherwise known as sciatica.
As another example, the presence of A-waves in multiple nerves of a
diabetic individual is indicative of diabetic polyneuropathy. As
yet another example, the occurrence of complex A-waves in a patient
presenting with rapid onset proximal weakness is the earliest sign
of Guillan-Barre syndrome.
* * * * *